Periodic communication logics for the decentralized control of multi-agent systems

This paper studies a system consisting of several dynamical subsystems (agents) that coordinate their actions by broadcasting their local state information over a communication network. It is assumed each agent can only observe its local state and that the medium-access control (MAC) protocol guarantees collision free broadcasts. Finally we assume that all agents have an internal dynamical model of their neighbors, so it is possible to estimate the neighboring agent's state between consecutive broadcasts from that agent. This paper examines open-loop communication logics that seek to minimize the entire group's aggregate state estimation error while minimizing the average broadcast rate. This paper's main result shows that open-loop communication logics requiring periodic broadcasts are optimal with respect to the aforementioned performance measure